How to analyze a graph and search for patterns using a neural network?

Question

Python

Hello! The task was to determine the features of the structure of electromagnet-crystalline inhomogeneity according to known characteristics using a neural network. That is, the structure is given in CST Microwave Studio, an example is shown under the spoiler. It is necessary, having its characteristic S11 (given under the second spoiler), using a neural network to determine the geometry, for example, the diameter of the hole on the back. How can this be done?
I plan to implement this in Python using the keras library. Is it rational? Maybe you should look towards matlab?

Structure

Characteristic S11

Answer 1

How can this be done?
I plan to implement this in Python using the keras library. Is it rational? Maybe you should look towards matlab?

It is necessary to look not at the implementation tool, but towards the design of the National Assembly.
I see that there are X,Y indicators for 5 different indicators (number of graphs) - these are input values ​​for training the NS.
Hole diameter (or something else you have ....) - output values .
According to the input and output data types and the process of generating output data based on input data - choose the desired type of neural network, design the required number of layers and their relationships. Adjust and test so that the NN works as you need and....
....ONLY AFTER ALL THIS - CODE.....

Answer 2

Vlob no. You need to think not about "neural networks", but about a model that adequately describes the processes.
When you understand the processes, you will understand where neural networks can be used - a universal approximator.